Hi,
On 25 Nov 2008, at 17:47, Jerry Yeou-Wei Chen wrote:
> Thanks again for the reply Steve.
> However, I could use some more clarification of two points:
>
> 1) For the .con file, does PPHeights refer to the range of values
> for each
> column of contrast matrices, or the design matrices (like PPheights
> in the
> .mat file)?
It's the peak-peak height of the "effective regressors" for each
contrast (see the NeuroImage design efficiency paper), however isn't
used by randomise so don't worry about this.
> 2) For the .con file again, does the number of contrasts factor into
> the
> statistics at all? (e.g. for correction for multiple comparisons) or
> does
> the "randomise" function just use it to know how many lines to read?
Well - people should really correct for multiple comparisons across
contrasts but no-one does.....randomise doesn't do this correction for
you.
Cheers.
>
>
> Thanks for your time,
> - Jerry
>
> On Thu, 13 Nov 2008 08:29:26 +0000, Steve Smith
> <[log in to unmask]> wrote:
>
>> Hi,
>>
>> Jolly good - just one minor comment below.
>>
>> Cheers, Steve.
>>
>>
>> On 12 Nov 2008, at 21:33, Jerry Yeou-Wei Chen wrote:
>>
>>> Hello,
>>>
>>> I would like to run a simple 2 sample t-test with a nuisance
>>> variable, and
>>> also run a linear regression on TBSS data.
>>> From what I've read in the manuals and the mail list, I think I have
>>> formulated the correct steps, but I would like to double check in
>>> particular
>>> the steps to create the design matrix and contrast files. Hopefully
>>> others
>>> can also benefit from having this info all together in one
>>> message! (I
>>> apologize beforehand for the long message! and any lack of clarity!)
>>>
>>> After running all the tbss scripts,
>>> For Simple 2 sample t-test:
>>> - create design matrix and contrast files using the design_ttest2
>>> command
>>>
>>> For Simple 2 sample t-test w/ 1 nuisance variable:
>>> - take the files from above and open them with a text editor
>>> - for the .mat file, add a new column of values under Matrix for
>>> your
>>> nuisance variable (mean-centred)
>>> - also change NumWaves to 3, and add a value to PPheights for the
>>> difference
>>> between the highest and lowest values of your nuisance variable
>>> - for the .con file, add a 0 to the end of each contrast
>>> - resulting tstat1 will show areas with higher FA in group 1, tstat2
>>> shows
>>> areas higher in group 2
>>>
>>> For regression
>>> - use the same method as above for creating the nuisance variable
>>> - if you are interested in interactions between group and a
>>> variable, split
>>> it to two regressors each containing variable values for one group
>>> and zeros
>>> for the other group
>>
>> Yes - but remember to demean each group's set of values before
>> padding
>> with zeros.
>>
>>> - for the .con file, use a 1 to test for positive slopes, -1 to test
>>> for
>>> negative slopes, and for interactions use a 1 and -1 for the two
>>> interaction
>>> regressors
>>> - resulting tstats will show areas with FA positively or negatively
>>> related
>>> to your regressor (for contrasts simply consisting of 1 or -1) or
>>> areas
>>> where one group's FA is related differently to your regressor (for
>>> contrasts
>>> with 1 and -1 for interaction regressors)
>>>
>>> Please let me know if there are any mistakes with this procedure!
>>>
>>> Thanks,
>>> - Jerry
>>>
>>
>>
>> ---------------------------------------------------------------------------
>> Stephen M. Smith, Professor of Biomedical Engineering
>> Associate Director, Oxford University FMRIB Centre
>>
>> FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
>> +44 (0) 1865 222726 (fax 222717)
>> [log in to unmask] http://www.fmrib.ox.ac.uk/~steve
>> ---------------------------------------------------------------------------
>
>
>
---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------
|